fedimovies/src/activitypub/handlers/create.rs
2023-01-29 20:04:23 +00:00

678 lines
23 KiB
Rust

use std::collections::HashMap;
use std::path::Path;
use chrono::Utc;
use serde_json::{Value as JsonValue};
use uuid::Uuid;
use crate::activitypub::{
constants::{AP_MEDIA_TYPE, AP_PUBLIC, AS_MEDIA_TYPE},
fetcher::fetchers::fetch_file,
fetcher::helpers::{
get_or_import_profile_by_actor_address,
get_or_import_profile_by_actor_id,
import_post,
},
identifiers::parse_local_actor_id,
receiver::{parse_array, parse_property_value, HandlerError},
types::{Attachment, EmojiTag, Link, LinkTag, Object, Tag},
vocabulary::*,
};
use crate::config::{Config, Instance};
use crate::database::{DatabaseClient, DatabaseError};
use crate::errors::{ConversionError, ValidationError};
use crate::models::{
attachments::queries::create_attachment,
emojis::queries::{
create_emoji,
get_emoji_by_remote_object_id,
update_emoji,
},
emojis::types::EmojiImage,
emojis::validators::{
validate_emoji_name,
EMOJI_MAX_SIZE,
EMOJI_MEDIA_TYPES,
},
posts::{
hashtags::normalize_hashtag,
helpers::get_post_by_object_id,
mentions::mention_to_address,
queries::create_post,
types::{Post, PostCreateData, Visibility},
validators::{
content_allowed_classes,
ATTACHMENTS_MAX_NUM,
CONTENT_MAX_SIZE,
EMOJIS_MAX_NUM,
},
},
profiles::types::DbActorProfile,
users::queries::get_user_by_name,
};
use crate::utils::{
html::clean_html,
urls::get_hostname,
};
use super::HandlerResult;
fn get_note_author_id(object: &Object) -> Result<String, ValidationError> {
let attributed_to = object.attributed_to.as_ref()
.ok_or(ValidationError("unattributed note"))?;
let author_id = parse_array(attributed_to)
.map_err(|_| ValidationError("invalid attributedTo property"))?
.get(0)
.ok_or(ValidationError("invalid attributedTo property"))?
.to_string();
Ok(author_id)
}
fn parse_object_url(value: &JsonValue) -> Result<String, ConversionError> {
let object_url = match value {
JsonValue::String(string) => string.to_owned(),
other_value => {
let links: Vec<Link> = parse_property_value(other_value)?;
if let Some(link) = links.get(0) {
link.href.clone()
} else {
return Err(ConversionError);
}
},
};
Ok(object_url)
}
pub fn get_note_content(object: &Object) -> Result<String, ValidationError> {
let mut content = if let Some(ref content) = object.content {
if object.media_type == Some("text/markdown".to_string()) {
format!("<p>{}</p>", content)
} else {
// HTML
content.to_string()
}
} else {
// Lemmy pages and PeerTube videos have "name" property
object.name.as_deref().unwrap_or("").to_string()
};
if object.object_type != NOTE {
// Append link to object
let object_url = if let Some(ref value) = object.url {
parse_object_url(value)
.map_err(|_| ValidationError("invalid object URL"))?
} else {
object.id.clone()
};
content += &format!(
r#"<p><a href="{0}">{0}</a></p>"#,
object_url,
);
};
if content.len() > CONTENT_MAX_SIZE {
return Err(ValidationError("content is too long"));
};
let content_safe = clean_html(&content, content_allowed_classes());
Ok(content_safe)
}
fn get_note_visibility(
author: &DbActorProfile,
primary_audience: Vec<String>,
secondary_audience: Vec<String>,
) -> Visibility {
let audience = [primary_audience, secondary_audience].concat();
// Some servers (e.g. Takahe) use "as" namespace
const PUBLIC_VARIANTS: [&str; 3] = [
AP_PUBLIC,
"as:Public",
"Public",
];
if audience.iter().any(|item| PUBLIC_VARIANTS.contains(&item.as_str())) {
return Visibility::Public;
};
let maybe_followers = author.actor_json.as_ref()
.and_then(|actor| actor.followers.as_ref());
if let Some(followers) = maybe_followers {
if audience.contains(followers) {
return Visibility::Followers;
};
};
let maybe_subscribers = author.actor_json.as_ref()
.and_then(|actor| actor.subscribers.as_ref());
if let Some(subscribers) = maybe_subscribers {
if audience.contains(subscribers) {
return Visibility::Subscribers;
};
};
Visibility::Direct
}
const ATTACHMENT_MAX_SIZE: usize = 20 * 1000 * 1000; // 20 MB
fn is_gnu_social_link(author_id: &str, attachment: &Attachment) -> bool {
if !author_id.contains("/index.php/user/") {
return false;
};
if attachment.attachment_type != DOCUMENT {
return false;
};
match attachment.media_type.as_ref() {
None => true,
Some(media_type) if media_type.contains("text/html") => true,
_ => false,
}
}
pub async fn handle_note(
db_client: &mut impl DatabaseClient,
instance: &Instance,
media_dir: &Path,
object: Object,
redirects: &HashMap<String, String>,
) -> Result<Post, HandlerError> {
match object.object_type.as_str() {
NOTE => (),
ARTICLE | EVENT | QUESTION | PAGE | VIDEO => {
log::info!("processing object of type {}", object.object_type);
},
other_type => {
log::warn!("discarding object of type {}", other_type);
return Err(ValidationError("unsupported object type").into());
},
};
let author_id = get_note_author_id(&object)?;
let author = get_or_import_profile_by_actor_id(
db_client,
instance,
media_dir,
&author_id,
).await.map_err(|err| {
log::warn!("failed to import {} ({})", author_id, err);
err
})?;
let content = get_note_content(&object)?;
let created_at = object.published.unwrap_or(Utc::now());
let mut attachments: Vec<Uuid> = Vec::new();
if let Some(value) = object.attachment {
let list: Vec<Attachment> = parse_property_value(&value)
.map_err(|_| ValidationError("invalid attachment property"))?;
let mut downloaded = vec![];
for attachment in list {
match attachment.attachment_type.as_str() {
DOCUMENT | IMAGE | VIDEO => (),
_ => {
log::warn!(
"skipping attachment of type {}",
attachment.attachment_type,
);
continue;
},
};
if is_gnu_social_link(&author_id, &attachment) {
// Don't fetch HTML pages attached by GNU Social
continue;
};
let attachment_url = attachment.url
.ok_or(ValidationError("attachment URL is missing"))?;
let (file_name, file_size, maybe_media_type) = fetch_file(
instance,
&attachment_url,
attachment.media_type.as_deref(),
ATTACHMENT_MAX_SIZE,
media_dir,
).await
.map_err(|err| {
log::warn!("{}", err);
ValidationError("failed to fetch attachment")
})?;
log::info!("downloaded attachment {}", attachment_url);
downloaded.push((file_name, file_size, maybe_media_type));
// Stop downloading if limit is reached
if downloaded.len() >= ATTACHMENTS_MAX_NUM {
log::warn!("too many attachments");
break;
};
};
for (file_name, file_size, maybe_media_type) in downloaded {
let db_attachment = create_attachment(
db_client,
&author.id,
file_name,
file_size,
maybe_media_type,
).await?;
attachments.push(db_attachment.id);
};
};
if content.is_empty() && attachments.is_empty() {
return Err(ValidationError("post is empty").into());
};
let mut mentions: Vec<Uuid> = Vec::new();
let mut hashtags = vec![];
let mut links = vec![];
let mut emojis = vec![];
if let Some(value) = object.tag {
let list: Vec<JsonValue> = parse_property_value(&value)
.map_err(|_| ValidationError("invalid tag property"))?;
for tag_value in list {
let tag_type = tag_value["type"].as_str().unwrap_or(HASHTAG);
if tag_type == HASHTAG {
let tag: Tag = match serde_json::from_value(tag_value) {
Ok(tag) => tag,
Err(_) => {
log::warn!("invalid hashtag");
continue;
},
};
if let Some(tag_name) = tag.name {
// Ignore invalid tags
if let Ok(tag_name) = normalize_hashtag(&tag_name) {
if !hashtags.contains(&tag_name) {
hashtags.push(tag_name);
};
};
};
} else if tag_type == MENTION {
let tag: Tag = match serde_json::from_value(tag_value) {
Ok(tag) => tag,
Err(_) => {
log::warn!("invalid mention");
continue;
},
};
// Try to find profile by actor ID.
if let Some(href) = tag.href {
if let Ok(username) = parse_local_actor_id(&instance.url(), &href) {
let user = get_user_by_name(db_client, &username).await?;
if !mentions.contains(&user.id) {
mentions.push(user.id);
};
continue;
};
// NOTE: `href` attribute is usually actor ID
// but also can be actor URL (profile link).
match get_or_import_profile_by_actor_id(
db_client,
instance,
media_dir,
&href,
).await {
Ok(profile) => {
if !mentions.contains(&profile.id) {
mentions.push(profile.id);
};
continue;
},
Err(error) => {
log::warn!(
"failed to find mentioned profile by ID {}: {}",
href,
error,
);
},
};
};
// Try to find profile by actor address
let tag_name = match tag.name {
Some(name) => name,
None => {
log::warn!("failed to parse mention");
continue;
},
};
if let Ok(actor_address) = mention_to_address(&tag_name) {
let profile = match get_or_import_profile_by_actor_address(
db_client,
instance,
media_dir,
&actor_address,
).await {
Ok(profile) => profile,
Err(error @ (
HandlerError::FetchError(_) |
HandlerError::DatabaseError(DatabaseError::NotFound(_))
)) => {
// Ignore mention if fetcher fails
// Ignore mention if local address is not valid
log::warn!(
"failed to find mentioned profile {}: {}",
actor_address,
error,
);
continue;
},
Err(other_error) => return Err(other_error),
};
if !mentions.contains(&profile.id) {
mentions.push(profile.id);
};
} else {
log::warn!("failed to parse mention {}", tag_name);
};
} else if tag_type == LINK {
let tag: LinkTag = match serde_json::from_value(tag_value) {
Ok(tag) => tag,
Err(_) => {
log::warn!("invalid link tag");
continue;
},
};
if tag.media_type != AP_MEDIA_TYPE &&
tag.media_type != AS_MEDIA_TYPE
{
// Unknown media type
continue;
};
let href = redirects.get(&tag.href).unwrap_or(&tag.href);
let linked = get_post_by_object_id(
db_client,
&instance.url(),
href,
).await?;
if !links.contains(&linked.id) {
links.push(linked.id);
};
} else if tag_type == EMOJI {
let tag: EmojiTag = match serde_json::from_value(tag_value) {
Ok(tag) => tag,
Err(error) => {
log::warn!("invalid emoji tag: {}", error);
continue;
},
};
if emojis.len() >= EMOJIS_MAX_NUM {
log::warn!("too many emojis");
continue;
};
let tag_name = tag.name.trim_matches(':');
if validate_emoji_name(tag_name).is_err() {
log::warn!("invalid emoji name");
continue;
};
let maybe_emoji_id = match get_emoji_by_remote_object_id(
db_client,
&tag.id,
).await {
Ok(emoji) => {
if emoji.updated_at >= tag.updated {
// Emoji already exists and is up to date
if !emojis.contains(&emoji.id) {
emojis.push(emoji.id);
};
continue;
};
if emoji.emoji_name != tag_name {
log::warn!("emoji name can't be changed");
continue;
};
Some(emoji.id)
},
Err(DatabaseError::NotFound("emoji")) => None,
Err(other_error) => return Err(other_error.into()),
};
let (file_name, file_size, maybe_media_type) = match fetch_file(
instance,
&tag.icon.url,
tag.icon.media_type.as_deref(),
EMOJI_MAX_SIZE,
media_dir,
).await {
Ok(file) => file,
Err(error) => {
log::warn!("failed to fetch emoji: {}", error);
continue;
},
};
let media_type = match maybe_media_type {
Some(media_type) if EMOJI_MEDIA_TYPES.contains(&media_type.as_str()) => {
media_type
},
_ => {
log::warn!(
"unexpected emoji media type: {:?}",
maybe_media_type,
);
continue;
},
};
log::info!("downloaded emoji {}", tag.icon.url);
let image = EmojiImage { file_name, file_size, media_type };
let emoji = if let Some(emoji_id) = maybe_emoji_id {
update_emoji(
db_client,
&emoji_id,
image,
&tag.updated,
).await?
} else {
let hostname = get_hostname(&tag.id)
.map_err(|_| ValidationError("invalid emoji ID"))?;
create_emoji(
db_client,
tag_name,
Some(&hostname),
image,
Some(&tag.id),
&tag.updated,
).await?
};
if !emojis.contains(&emoji.id) {
emojis.push(emoji.id);
};
} else {
log::warn!(
"skipping tag of type {}",
tag_type,
);
};
};
};
if let Some(ref object_id) = object.quote_url {
let object_id = redirects.get(object_id).unwrap_or(object_id);
let linked = get_post_by_object_id(
db_client,
&instance.url(),
object_id,
).await?;
if !links.contains(&linked.id) {
links.push(linked.id);
};
};
let in_reply_to_id = match object.in_reply_to {
Some(ref object_id) => {
let object_id = redirects.get(object_id).unwrap_or(object_id);
let in_reply_to = get_post_by_object_id(
db_client,
&instance.url(),
object_id,
).await?;
Some(in_reply_to.id)
},
None => None,
};
let primary_audience = match object.to {
Some(value) => {
parse_array(&value)
.map_err(|_| ValidationError("invalid 'to' property value"))?
},
None => vec![],
};
let secondary_audience = match object.cc {
Some(value) => {
parse_array(&value)
.map_err(|_| ValidationError("invalid 'cc' property value"))?
},
None => vec![],
};
let visibility = get_note_visibility(
&author,
primary_audience,
secondary_audience,
);
if visibility != Visibility::Public {
log::warn!(
"processing note with visibility {:?} attributed to {}",
visibility,
author.username,
);
};
let post_data = PostCreateData {
content: content,
in_reply_to_id,
repost_of_id: None,
visibility,
attachments: attachments,
mentions: mentions,
tags: hashtags,
links: links,
emojis: emojis,
object_id: Some(object.id),
created_at,
};
let post = create_post(db_client, &author.id, post_data).await?;
Ok(post)
}
pub async fn handle_create(
config: &Config,
db_client: &mut impl DatabaseClient,
activity: JsonValue,
is_authenticated: bool,
) -> HandlerResult {
let object: Object = serde_json::from_value(activity["object"].to_owned())
.map_err(|_| ValidationError("invalid object"))?;
let object_id = object.id.clone();
let object_received = if is_authenticated {
Some(object)
} else {
// Fetch object, don't trust the sender.
// Most likely it's a forwarded reply.
None
};
import_post(config, db_client, object_id, object_received).await?;
Ok(Some(NOTE))
}
#[cfg(test)]
mod tests {
use serde_json::json;
use crate::activitypub::{
actors::types::Actor,
types::Object,
vocabulary::NOTE,
};
use super::*;
#[test]
fn test_get_note_author_id() {
let object = Object {
object_type: NOTE.to_string(),
attributed_to: Some(json!(["https://example.org/1"])),
..Default::default()
};
let author_id = get_note_author_id(&object).unwrap();
assert_eq!(author_id, "https://example.org/1");
}
#[test]
fn test_get_note_content() {
let object = Object {
content: Some("test".to_string()),
object_type: NOTE.to_string(),
..Default::default()
};
let content = get_note_content(&object).unwrap();
assert_eq!(content, "test");
}
#[test]
fn test_get_note_content_from_video() {
let object = Object {
name: Some("test-name".to_string()),
content: Some("test-content".to_string()),
object_type: "Video".to_string(),
url: Some(json!([{
"type": "Link",
"mediaType": "text/html",
"href": "https://example.org/xyz",
}])),
..Default::default()
};
let content = get_note_content(&object).unwrap();
assert_eq!(
content,
r#"test-content<p><a href="https://example.org/xyz" rel="noopener">https://example.org/xyz</a></p>"#,
);
}
#[test]
fn test_get_note_visibility_public() {
let author = DbActorProfile::default();
let primary_audience = vec![AP_PUBLIC.to_string()];
let secondary_audience = vec![];
let visibility = get_note_visibility(
&author,
primary_audience,
secondary_audience,
);
assert_eq!(visibility, Visibility::Public);
}
#[test]
fn test_get_note_visibility_followers() {
let author_followers = "https://example.com/users/author/followers";
let author = DbActorProfile {
actor_json: Some(Actor {
followers: Some(author_followers.to_string()),
..Default::default()
}),
..Default::default()
};
let primary_audience = vec![author_followers.to_string()];
let secondary_audience = vec![];
let visibility = get_note_visibility(
&author,
primary_audience,
secondary_audience,
);
assert_eq!(visibility, Visibility::Followers);
}
#[test]
fn test_get_note_visibility_subscribers() {
let author_followers = "https://example.com/users/author/followers";
let author_subscribers = "https://example.com/users/author/subscribers";
let author = DbActorProfile {
actor_json: Some(Actor {
followers: Some(author_followers.to_string()),
subscribers: Some(author_subscribers.to_string()),
..Default::default()
}),
..Default::default()
};
let primary_audience = vec![author_subscribers.to_string()];
let secondary_audience = vec![];
let visibility = get_note_visibility(
&author,
primary_audience,
secondary_audience,
);
assert_eq!(visibility, Visibility::Subscribers);
}
#[test]
fn test_get_note_visibility_direct() {
let author = DbActorProfile::default();
let primary_audience = vec!["https://example.com/users/1".to_string()];
let secondary_audience = vec![];
let visibility = get_note_visibility(
&author,
primary_audience,
secondary_audience,
);
assert_eq!(visibility, Visibility::Direct);
}
}